CN101901350B - Characteristic vector-based static gesture recognition method - Google Patents
Characteristic vector-based static gesture recognition method Download PDFInfo
- Publication number
- CN101901350B CN101901350B CN2010102383683A CN201010238368A CN101901350B CN 101901350 B CN101901350 B CN 101901350B CN 2010102383683 A CN2010102383683 A CN 2010102383683A CN 201010238368 A CN201010238368 A CN 201010238368A CN 101901350 B CN101901350 B CN 101901350B
- Authority
- CN
- China
- Prior art keywords
- gesture
- point
- vector
- angle
- sigma
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- User Interface Of Digital Computer (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a characteristic vector-based static gesture recognition method, which comprises the following steps: carrying out segmentation processing on a gesture by using color characteristics in an YCbCr color space; selecting a characteristic vector to describe the gesture, and carrying out length and angle normalization processing; determining the position of a fingertip in the gesture according to the characteristic vector; and finally recognizing the gesture by using the relative positions of the fingertip and the gesture center. The method has the advantages of rotational invariance, scale invariance and good real-time, and can meet the requirement for human-computer interaction application.
Description
(1) technical field:
The present invention relates to a kind of static gesture identification method, especially a kind of static gesture identification method based on proper vector.Belong to field of human-computer interaction.
(2) background technology:
Along with the development of human-computer interaction technology, the man-machine interaction mode of natural harmony receives people's attention day by day.Based on the Gesture Recognition of vision, with its naturality, terseness and direct characteristics, the Man Machine Interface of a nature is provided, received researcher's attention.Because the characteristic of the non-rigid body of staff; Traditional recognition methods with rotational invariance and convergent-divergent unchangeability; For example fourier descriptors, Image Moment Invariants etc. have run into variety of problems in the application of gesture identification, can not satisfy the requirement of man-machine interaction to the gesture identification real-time.
Through the existing literature source investigation is found; In the paper " based on the gesture identification method of fourier descriptors-BP neural network " that Qin Wenjun etc. 2009 delivered in " Northeastern University's journal (natural science edition) " the 30th the 9th phase of volume a kind of gesture identification method based on fourier descriptors and BP neural network has been proposed; At first the gesture dividing method through many Feature Fusion extracts the gesture zone; Combine then fourier descriptors preferably profile describe ability and BP neural network self-learning capability preferably, with fourier descriptors and BP neural network method gesture is discerned.The method of this paper has obtained high recognition, but the algorithm computation amount is big, and the identification piece image needs 4.5 second time, far can not reach the requirement of real-time application; Because fourier descriptors is applicable to the profile description to rigid body, therefore be applied in the gesture identification simultaneously, can not well describe the same gesture under the different angles, the poor robustness of its algorithm.
(3) summary of the invention:
The objective of the invention is to propose a kind of static gesture identification method based on proper vector, prior art calculated amount in gesture identification is used is big to solve, recognition result receives the big problem of gesture aspect effect, overcomes defective of the prior art.
Technical scheme of the present invention is summarised as: at first in the YCbCr color space, with the colour of skin characteristic gesture is carried out dividing processing; Selected characteristic vector description gesture then; Carrying out length and angle normalization handles; Confirm the position of finger tip in the gesture finally gesture to be discerned according to proper vector with the relative position at finger tip and gesture center.
A kind of static gesture identification method of the present invention based on proper vector, its concrete steps are following:
Step 1: gesture is cut apart
Adopt Logitech QuickCam Pro9000 camera collection image, obtaining picture size is 320 * 240, and picture format is JPEG, and shooting background does not have the zone of the similar colour of skin, and the user who is taken wears non-colour of skin caftan.To the image that obtains, at first be transformed into the YCbCr color space from the RGB color space, select the central value and the threshold value of Cb, Cr component afterwards, wherein Cb component central value Cb
Mid=115, Cr component central value Cr
Mid=145, threshold value l
Thres=15, each pixel of input picture is calculated the Euclidean distance dist of its Cb, Cr value and central value:
If dist is less than corresponding threshold value l
Thres, then the value with this pixel is made as 1, otherwise its value is made as 0, and input picture is converted into bianry image.Bianry image to obtaining once corrodes expansion process, and largest connected zone is the gesture region in the supposition image.
Step 2: the center, direction and the edge that obtain gesture
After images of gestures is converted into bianry image, through calculating center, the direction that obtains gesture, and extract the edge of gesture.Concrete grammar is following:
Calculate gesture center point coordinate x and y by formula (2) and formula (3):
In the formula, the value of the capable j row of i in B [i, the j] presentation video, the size of image is n * m.
The calculating formula of gesture direction θ is following:
In the formula, μ
11, μ
20, μ
02Second-order moment around mean for image.
From bianry image, extract the edge of gesture with the Canny operator.
Step 3: selected characteristic vector
Calculating the angle of gesture central point and edge each point line, find out the immediate point of angle and gesture deflection, is starting point with this point, and the n five equilibrium is carried out on the gesture border, gets borderline n point.With the gesture central point is starting point, and one of borderline n point characterizes this gesture for terminating point structure vector with n vector.Be the influence of size of eliminating gesture in the input picture to discerning, the longest vector of length in the selected characteristic vector, its length is set to 100, and the length normalization method that other is vectorial is to (0,100).Through experiment, adopt n=60.
Step 4: location fingertip location
The proper vector that step 3 is obtained is regarded as an end to end Vector Groups, eliminates starting point and selects the influence to Vector Groups, guarantees that proper vector is insensitive to rotating.From Vector Groups, obtain local maximum and local minimum, the point that wherein local maximum is corresponding is a finger tip point candidate point, and the differential seat angle of its two adjacent local minizing points is the angle of finger region.Selected finger length threshold value l
ThWith angle threshold θ
Th, use finger tip to compare with setting threshold with the angle θ that points the region and judge whether the finger tip candidate point is finger tip to the length l at gesture center.As l>l
ThAnd θ<θ
ThThe time, then the terminal point position of this proper vector is a fingertip location, otherwise this is non-finger tip.Through experiment, adopt l
Th=53, θ
Th=1.87.
Step 5: gesture identification
Obtain gesture central point and fingertip location according to step 2 and step 4, constructing one group is starting point with the gesture central point, and fingertip location is the vector of terminal point.Maximum angle between finding out in these vectors in twos; With one in two vectors of angle maximum is start vector; Its angle is set to 0; Guaranteeing under the constant prerequisite of each vectorial relative angle, making other vector minimum, obtaining unique normalization result with the counterclockwise angle and the sum of start vector.Obtain confirming the finger number that gesture comprises after the normalized Vector Groups of angle; According to the relative position and the length information of finger, the different gestures that comprise the same hand index are distinguished.
Advantage of the present invention and effect are: the identification to gesture has rotational invariance and convergent-divergent unchangeability, calculates simply, satisfies real-time processing requirements, is applicable to the identification of the different gestures of different staff.
(4) description of drawings:
Fig. 1 is the algorithm block diagram of the inventive method.
Fig. 2 is the center and the direction synoptic diagram of gesture, and rhombus is represented the center of gesture among the figure, and arrow is represented the direction of gesture among the figure.
Fig. 3 is the proper vector synoptic diagram of gesture.
Wherein, Fig. 3 (a) is 2 finger gesture proper vector synoptic diagram;
Fig. 3 (b) is 4 finger gesture proper vector synoptic diagram.
Fig. 4 is fingertip location figure, and rhombus is the fingertip location of the inventive method location among the figure.
Wherein, Fig. 4 (a) is 3 finger locating figure as a result;
Fig. 4 (b) is 5 finger locating figure as a result;
(5) practical implementation method:
Technical scheme of the present invention is summarised as: at first in the YCbCr color space, with the colour of skin characteristic gesture is carried out dividing processing; Selected characteristic vector description gesture then; Carrying out length and angle normalization handles; Confirm the position of finger tip in the gesture finally gesture to be discerned according to proper vector with the relative position at finger tip and gesture center.
Below in conjunction with accompanying drawing technical scheme of the present invention is done further to describe in detail.Key step is following:
Step 1: gesture is cut apart
The present invention adopts Logitech QuickCam Pro9000 camera collection image, obtains coloured image and is of a size of 320 * 240, and picture format is JPEG, and shooting background does not have the zone of the similar colour of skin, and the user who is taken wears non-colour of skin caftan.To the images of gestures that obtains, at first be transformed into the YCbCr color space from the RGB color space, select the central value and the threshold value of Cb, Cr component afterwards, wherein Cb component central value Cb
Mid=115, Cr component central value Cr
Mid=145, threshold value l
Thres=15, each pixel of input picture is calculated the Euclidean distance dist of its Cb, Cr value and central value:
If dist is less than corresponding threshold value l
Thres, then this pixel value is made as 1, otherwise its value is made as 0, thereby input picture is converted into bianry image.Bianry image to obtaining once corrodes expansion process, and largest connected zone is the gesture region in the supposition image.
Step 2: the center, direction and the edge that obtain gesture
After images of gestures is converted into bianry image, through calculating center, the direction that obtains gesture, and extract the edge of gesture.
Calculate gesture center point coordinate x and y by formula (2) and formula (3):
In the formula, the value of the capable j row of i in B [i, the j] presentation video, the size of image is n * m.
The calculating formula of gesture direction θ is following:
In the formula, μ
11, μ
20, μ
02Second-order moment around mean for image.
From bianry image, extract the edge of gesture with the Canny operator.
Fig. 2 is the center and the direction synoptic diagram of gesture, and rhombus is represented the center of gesture among the figure, and arrow is represented the direction of gesture among the figure.
Step 3: selected characteristic vector
Calculating the angle of gesture central point and edge each point line, find out the immediate point of angle and gesture deflection, is starting point with this point, and the n five equilibrium is carried out on the gesture border, gets borderline n point.With the gesture central point is starting point, and one of borderline n point characterizes this gesture for terminating point structure vector with n vector.Be the influence of size of eliminating gesture in the input picture to discerning, the longest vector of length in the selected characteristic vector, its length is set to 100, and the length normalization method that other is vectorial is to [0,100].Through experiment, adopt n=60.
Fig. 3 (a) is to 2 finger gesture selected characteristic vector result figure, and Fig. 3 (b) is to 4 finger gesture selected characteristic vector result figure,
Step 4: location fingertip location
The proper vector that step 3 is obtained is regarded as an end to end Vector Groups, eliminates starting point and selects the influence to Vector Groups, guarantees that proper vector is insensitive to rotating.From Vector Groups, obtain local maximum and local minimum, the point that wherein local maximum is corresponding is a finger tip point candidate point, and the differential seat angle of its two adjacent local minizing points is the angle of finger region.Selected finger length threshold value l
ThWith angle threshold θ
Th, use finger tip to compare with setting threshold with the angle θ that points the region and judge whether the finger tip candidate point is finger tip to the length l at gesture center.As l>l
ThAnd θ<θ
ThThe time, then the terminal point position of this proper vector is a fingertip location, otherwise this is non-finger tip.Through experiment, adopt l
Th=53, θ
Th=1.87.
Fig. 4 (a) is the finger tip positioning result figure to 3 finger gestures, and Fig. 4 (b) is the finger tip positioning result figure to 5 finger gestures.
Step 5: gesture identification
Obtain gesture central point and fingertip location according to step 2 and step 4, constructing one group is starting point with the gesture central point, and fingertip location is the vector of terminal point.Maximum angle between finding out in these vectors in twos; With one in two vectors of angle maximum is start vector; Its angle is set to 0; Guaranteeing under the constant prerequisite of each vectorial relative angle, making other vector minimum, obtaining unique normalization result with the counterclockwise angle and the sum of start vector.Obtain having confirmed the finger number that gesture comprises after the normalized Vector Groups of angle; According to the relative position and the length information of finger, the different gestures that comprise the same hand index are distinguished.The inventive method was discerned a width of cloth images of gestures about 0.4 second consuming time, and that the method for Qin Wenjun etc. is discerned a width of cloth images of gestures is consuming time about 4.5 seconds.Be in advantage of the present invention and effect: the identification to gesture has rotational invariance and convergent-divergent unchangeability, calculates simply, satisfies real-time processing requirements, and good discrimination is arranged.
The English symbol that occurs in the Figure of description, its implication is following:
L: with the length of finger tip to the gesture center;
θ: the angle of finger region;
l
Th: the finger length threshold value is made as 53;
θ
Th: angle threshold is made as 1.87.
Claims (1)
1. the static gesture identification method based on proper vector is characterized in that, comprises following several steps:
Step 1: gesture is cut apart
Adopt Logitech QuickCam Pro9000 camera collection image, obtaining picture size is 320 * 240, and picture format is JPEG, and shooting background does not have the zone of the similar colour of skin, and the user who is taken wears non-colour of skin caftan; To the image that obtains, at first be transformed into the YCbCr color space from the RGB color space, select the central value and the threshold value of Cb, Cr component afterwards, wherein Cb component central value Cb
Mid=115, Cr component central value Cr
Mid=145, threshold value l
Thres=15, each pixel of input picture is calculated the Euclidean distance dist of its Cb, Cr value and central value:
If dist is less than corresponding threshold value l
Thres, then the value with this pixel is made as 1, otherwise its value is made as 0, thereby input picture is converted into bianry image; Bianry image to obtaining once corrodes expansion process, and largest connected zone is the gesture region in the supposition image;
Step 2: the center, direction and the edge that obtain gesture
After images of gestures is converted into bianry image, through calculating center, the direction that obtains gesture, and extract the edge of gesture;
Calculate gesture center point coordinate x and y by formula (2) and formula (3):
In the formula, the value of the capable j row of i in B [i, the j] presentation video, the size of image is n * m;
The calculating formula of gesture direction θ is following:
In the formula, μ
11, μ
20, μ
02Second-order moment around mean for image;
From bianry image, extract the edge of gesture with the Canny operator;
Step 3: selected characteristic vector
Calculating the angle of gesture central point and edge each point line, find out the immediate point of angle and gesture deflection, is starting point with this point, and the n five equilibrium is carried out on the gesture border, obtains borderline n point; With the gesture central point is starting point, and one of borderline n point characterizes this gesture for terminating point structure vector with this n vector; Be the influence of size of eliminating gesture in the input picture to discerning, the longest vector of length in the selected characteristic vector, its length is set to 100, and the length normalization method that other is vectorial is to (0,100); Adopt n=60;
Step 4: location fingertip location
The proper vector that step 3 is obtained is regarded as an end to end Vector Groups, eliminates starting point and selects the influence to Vector Groups, guarantees that proper vector is insensitive to rotating; From Vector Groups, obtain local maximum and local minimum, the point that wherein local maximum is corresponding is a finger tip point candidate point, and the differential seat angle of its two adjacent local minizing points is the angle of finger region; Selected finger length threshold value l
ThWith angle threshold θ
Th, use finger tip to compare with setting threshold with the angle θ that points the region and judge whether the finger tip candidate point is finger tip to the length l at gesture center; As l>l
ThAnd θ<θ
ThThe time, then the terminal point position of this proper vector is a fingertip location, otherwise this is non-finger tip; Adopt l
Th=53, θ
Th=1.87;
Step 5: gesture identification
Obtain gesture central point and fingertip location according to step 2 and step 4, constructing one group is starting point with the gesture central point, and fingertip location is the vector of terminal point; Maximum angle between finding out in these vectors in twos; With one in two vectors of angle maximum is start vector; Its angle is set to 0; Guaranteeing under the constant prerequisite of each vectorial relative angle, making other vector minimum, obtaining unique normalization result with the counterclockwise angle and the sum of start vector; After obtaining the normalized Vector Groups of angle, then confirmed the finger number that gesture comprises; According to the relative position and the length information of finger, the different gestures that comprise the same hand index are distinguished.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010102383683A CN101901350B (en) | 2010-07-23 | 2010-07-23 | Characteristic vector-based static gesture recognition method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2010102383683A CN101901350B (en) | 2010-07-23 | 2010-07-23 | Characteristic vector-based static gesture recognition method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101901350A CN101901350A (en) | 2010-12-01 |
CN101901350B true CN101901350B (en) | 2012-05-16 |
Family
ID=43226873
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2010102383683A Expired - Fee Related CN101901350B (en) | 2010-07-23 | 2010-07-23 | Characteristic vector-based static gesture recognition method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101901350B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102938060A (en) * | 2012-12-07 | 2013-02-20 | 上海电机学院 | Dynamic gesture recognition system and method |
Families Citing this family (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102063618B (en) * | 2011-01-13 | 2012-10-31 | 中科芯集成电路股份有限公司 | Dynamic gesture identification method in interactive system |
CN103562822A (en) * | 2011-04-28 | 2014-02-05 | Nec软件系统科技有限公司 | Information processing device, information processing method, and recording medium |
US8929612B2 (en) * | 2011-06-06 | 2015-01-06 | Microsoft Corporation | System for recognizing an open or closed hand |
CN102855461B (en) * | 2011-07-01 | 2016-05-25 | 株式会社理光 | In image, detect the method and apparatus of finger |
CN102368290B (en) * | 2011-09-02 | 2012-12-26 | 华南理工大学 | Hand gesture identification method based on finger advanced characteristic |
CN103049446B (en) * | 2011-10-13 | 2016-01-27 | 中国移动通信集团公司 | A kind of image search method and device |
CN103778407A (en) * | 2012-10-23 | 2014-05-07 | 南开大学 | Gesture recognition algorithm based on conditional random fields under transfer learning framework |
WO2014080487A1 (en) * | 2012-11-22 | 2014-05-30 | 富士通株式会社 | Information processing device, body part determination program and body part determination method |
CN103544716B (en) * | 2013-10-30 | 2016-04-27 | 京北方信息技术股份有限公司 | A kind of pixel to image carries out method and the device of color classification |
CN104102904B (en) * | 2014-07-14 | 2016-03-23 | 济南大学 | A kind of static gesture identification method |
CN104751141A (en) * | 2015-03-30 | 2015-07-01 | 东南大学 | ELM gesture recognition algorithm based on feature image full pixel gray values |
CN105405143B (en) * | 2015-11-13 | 2017-04-12 | 华中师范大学 | Gesture segmentation method and system based on global expectation-maximization algorithm |
CN105867630A (en) * | 2016-04-21 | 2016-08-17 | 深圳前海勇艺达机器人有限公司 | Robot gesture recognition method and device and robot system |
CN106408579B (en) * | 2016-10-25 | 2019-01-29 | 华南理工大学 | A kind of kneading finger tip tracking based on video |
CN107702161A (en) * | 2017-08-23 | 2018-02-16 | 信利光电股份有限公司 | A kind of electromagnetic oven control method, device and electromagnetic oven |
CN109840478B (en) * | 2019-01-04 | 2021-07-02 | 广东智媒云图科技股份有限公司 | Action evaluation method and device, mobile terminal and readable storage medium |
CN110020634B (en) * | 2019-04-15 | 2020-05-19 | 刘政操 | Data show board for business administration |
CN114842712B (en) * | 2022-04-12 | 2023-10-17 | 汕头大学 | Sign language teaching system based on gesture recognition |
CN116884078B (en) * | 2023-09-07 | 2023-11-24 | 鹿客科技(北京)股份有限公司 | Image pickup apparatus control method, monitoring device, and computer-readable medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101685343A (en) * | 2008-09-26 | 2010-03-31 | 联想(北京)有限公司 | Method, device and electronic aid for realizing gesture identification |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2407635B (en) * | 2003-10-31 | 2006-07-12 | Hewlett Packard Development Co | Improvements in and relating to camera control |
-
2010
- 2010-07-23 CN CN2010102383683A patent/CN101901350B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101685343A (en) * | 2008-09-26 | 2010-03-31 | 联想(北京)有限公司 | Method, device and electronic aid for realizing gesture identification |
Non-Patent Citations (3)
Title |
---|
孙丽娟 等.基于视觉的手势识别技术.《计算机技术与发展》.2008,第18卷(第10期),214-216,221. * |
曹陶科 等.基于视觉的手势识别研究.《郑州大学学报(理学版)》.2008,第40卷(第3期),63-66. * |
覃文军 等.基于傅立叶描述子-BP神经网络的手势识别方法.《东北大学学报(自然科学版)》.2009,第30卷(第9期),1232-1235. * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102938060A (en) * | 2012-12-07 | 2013-02-20 | 上海电机学院 | Dynamic gesture recognition system and method |
Also Published As
Publication number | Publication date |
---|---|
CN101901350A (en) | 2010-12-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101901350B (en) | Characteristic vector-based static gesture recognition method | |
Dominio et al. | Combining multiple depth-based descriptors for hand gesture recognition | |
CN102368290B (en) | Hand gesture identification method based on finger advanced characteristic | |
Wang et al. | A new face detection method based on shape information | |
US8391615B2 (en) | Image recognition algorithm, method of identifying a target image using same, and method of selecting data for transmission to a portable electronic device | |
CN102231093B (en) | Screen locating control method and device | |
CN102063618B (en) | Dynamic gesture identification method in interactive system | |
CN107885327B (en) | Fingertip detection method based on Kinect depth information | |
Feng et al. | Features extraction from hand images based on new detection operators | |
CN103383700B (en) | Based on the edge direction histogrammic image search method of difference | |
CN105160323B (en) | A kind of gesture identification method | |
CN108304819B (en) | Gesture recognition system and method, and storage medium | |
CN101807257A (en) | Method for identifying information of image tag | |
CN103294996A (en) | 3D gesture recognition method | |
CN106326860A (en) | Gesture recognition method based on vision | |
CN101447021A (en) | Face fast recognition system and recognition method thereof | |
CN103455794A (en) | Dynamic gesture recognition method based on frame fusion technology | |
CN103530892A (en) | Kinect sensor based two-hand tracking method and device | |
CN109359566B (en) | Gesture recognition method for hierarchical classification by using finger characteristics | |
CN105760472A (en) | Video retrieval method and system | |
CN103839042A (en) | Human face recognition method and human face recognition system | |
CN103257713A (en) | Gesture control method | |
CN105975906B (en) | A kind of PCA static gesture identification methods based on area features | |
CN112906550A (en) | Static gesture recognition method based on watershed transformation | |
CN108022243A (en) | Method for detecting paper in a kind of image based on deep learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120516 Termination date: 20120723 |